CHARACTERISTICS OF DATA MINING BY CLASSIFICATION EDUCATIONAL DATASET TO IMPROVE STUDENT'S EVALUATION

The educational data mining (EDM) can be specified as one of the main fields related to high-quality research that involves mining datasets to address research questions related to education; such questions examine the ways in which people learn and teach. a large amount of data, including education data, are being collected, and much of them are unprocessed. The success of EDM was examined in this paper, and nine data mining techniques were explored including: bagging, multilayer perception (MLP), naive Bayes (NB), K-nearest neighbours (KNN), logistic regression (LR), support vector machine (SVM), XGBoost, decision tree (DT), and random forest (RF). Such techniques were used on an educational dataset obtained from certain learning management system which is referred to as Kalboard 360. This paper involves three major steps. Firstly, student performance model that contains exceptional feature's category, that are referred to as behavioural features, is introduced. Secondly, the dataset is pre-processed, and the pre-processing steps involve transforming the raw data into a usable format and verifying the connections between independent and dependent variables in sample dataset, which has been also referred to as the training dataset. Thirdly, the nine data mining approaches have been utilized on the acquired dataset to classify student performance into low, middle, and high levels. Afterwards, the performance measures were examined by using recall, precision, accuracy, as well as F1 score. RF (89%) obtained the best accuracy also other techniques were ordered in terms of accuracy: bagging (85%) > XGBoost 84% > NB (81%) > LR (81%) > MLP (77%) > DT (76%) > SVM (72%) > KNN (68%). Results were compared by using divided datasets (80:20 ratio) (80 for training: 20 for testing).